Cite this Article
Chen Weizhong, Huang Yao, He Lijun, Han Zhengsheng, Huang Yi. A snapback-free TOL-RC-LIGBT with vertical P-collector and N-buffer design. Chinese Physics B, 2018, 27(8): 088501  
Permissions
A snapback-free TOL-RC-LIGBT with vertical P-collector and N-buffer design
Chen Weizhong1, 2, †, Huang Yao1, ‡, He Lijun1, Han Zhengsheng2, 3, Huang Yi1
College of Electronics Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100029, China
University of Chinese Academy of Sciences, Beijing 100049, China

 

† Corresponding author. E-mail: cwz@cqu.edu.cn 632752486@qq.com

Project supported by the National Natural Science Foundation of China (Grant No. 61604027), the Basic and Advanced Technology Research Project of Chongqing Municipality, China (Grant No. cstc2016jcyjA1923), the Scientific and Technological Research Foundation of Chongqing Municipal Education Commission, China (Grant No. KJ1500404), the Youth Natural Science Foundation of Chongqing University of Posts and Telecommunications, China (Grant Nos. A2015-50 and A2015-52), the Chongqing Key Laboratory Improvement Plan, China (Chongqing Key Laboratory of Photo Electronic Information Sensing and Transmitting Technology) (Grant No. cstc2014pt-sy40001), and the University Innovation Team Construction Plan Funding Project of Chongqing, China (Architecture and Core Technologies of Smart Medical System) (Grant No. CXTDG201602009).

Abstract

A reverse-conducting lateral insulated-gate bipolar transistor (RC-LIGBT) with a trench oxide layer (TOL), featuring a vertical N-buffer and P-collector is proposed. Firstly, the TOL enhances both of the surface and bulk electric fields of the N-drift region, thus the breakdown voltage (BV) is improved. Secondly, the vertical N-buffer layer increases the voltage drop VPN of the P-collector/N-buffer junction, thus the snapback is suppressed. Thirdly, the P-body and the vertical N-buffer act as the anode and the cathode, respectively, to conduct the reverse current, thus the inner diode is integrated. As shown by the simulation results, the proposed RC-LIGBT exhibits trapezoidal electric field distribution with BV of 342.4 V, which is increased by nearly 340% compared to the conventional RC-LIGBT with triangular electric fields of 100.2 V. Moreover, the snapback is eliminated by the vertical N-buffer layer design, thus the reliability of the device is improved.

PACS: 85.30.De;85.30.Pq;85.30.Tv
Keyword:reverse-conducting lateral insulated-gate bipolar transistor (RC-LIGBT);breakdown voltage;snapback phenomenon
1. Introduction

Lateral insulated gate bipolar transistor (LIGBT) is widely applied to power integrated circuits (ICs) due to its voltage control and low power dissipation.[1–3] On the one hand, the breakdown voltage (BV) is a key property for the LIGBT and it is improved by the trench oxide layer (TOL) technology.[4–6] On the other hand, the reverse conduction capability is another key property for the LIGBT, and the reverse conducting LIGBT (RC-LIGBT) is introduced to integrate the IGBT and the free-wheeling diode (FWD) into one monolithic chip.[7,8] However, the snapback problem is observed due to the short effects of the N-collector at the forward conduction state.[9–12] In recent years, several advanced structures have been proposed to solve this issue.[13–18] The snapback mechanisms and models were given by the voltage drop VPN of the P-collector/N-buffer junction.[19,20] The VPN model is an effective way to suppress the snapback by increasing the doping of the N-drift, but the breakdown voltage will be decreased due to the electric field. On the other hand, the snapback can be suppressed by increasing the resistance of the lateral N-buffer layer. However, it also has adverse influence on the breakdown voltage because the N-buffer acts as the electric field stop layer. Therefore, increasing the breakdown voltage and suppressing the snapback contradict each other.

In this paper, a novel RC-LIGBT with TOL inserted in the N-drift and a vertical N-buffer layer sandwiched between the TOL and P-collector is proposed to improve the breakdown voltage at the reverse breakdown state and eliminate the snapback at the forward conduction state.

2. Device structures and key parameters

Figure 1 shows the structures and key parameters for the devices. Figure 1(a) shows the conventional RC-LIGBT with lateral N-buffer and P-collector in the collector. Figure 1(b) shows the proposed RC-LIGBT with TOL inserted in the N-drift and a vertical N-buffer layer sandwiched between the TOL and the P-collector. A PN junction is formed between the P-collector and the N-buffer for both devices. The two devices are designed with the same parameters except the N-buffer and the P-collector. For both devices, the doping of the N-drift region is 5 × 1014 cm−3, the thickness and the width of the N-drift region are 25 μm and 17 μm, and the thickness of the buried oxide (BOX) and the doping of the P-substrate are 2 μm and 8 × 1014 cm−3, respectively. Additionally, the thickness and the width of the TOL are 22 μm and 10 μm for the proposed RC-LIGBT, and the width of the vertical N-buffer Wn and the P-collector Wp can be adjusted. The electric characteristics of the devices are investigated by the MEDICI,[21] and the physical and electrical models such as mobility, ionization, and the recombination models are adopted.

Fig. 1. (color online) Structures and key parameters of (a) conventional RC-LIGBT with lateral N-buffer in the collector, and (b) proposed RC-LIGBT with trench oxide layer and vertical N-buffer in the collector.
3. Results and discussion
3.1. Reverse breakdown voltage BV

At the breakdown state, the gate electrode and the emitter electrode of the device are connected to the ground. VGate = 0 V, VEmitter = 0 V, and the collector electrode VCollector is applied with positive voltage. The collector current and the voltage curves are shown in Fig. 2 for the RC-LIGBTs. The doping of the N-drift Ndrift determines the slope of the electric field, and BV will increases with the decrease of Ndrift. The simulation shows that when Ndrift decreases from 1 × 1015 cm−3 to 3 × 1014 cm−3, BV increases from 238 V to 342.4 V for the proposed RC-LIGBT, and from 73.7 V to 100.2 V for the conventional RC-LIGBT. Thus the BV of the proposed RC-LIGBT is improved by 340% compared to that of the conventional RC-LIGBT at Ndrift = 3 × 1014 cm−3. Moreover, device A without TOL and device B with a right vertical N-buffer layer (illustrated in Fig. 3) are also investigated. Device A has the BV of 102 V and device B has the BV of 230 V.

Fig. 2. (color online) Comparison of the breakdown characteristics of the devices as the N-drift doping changes from 1 × 1015 cm−3 to 3 × 1014 cm−3. Device A and device B are illustrated in Fig. 3.

Figure 3 shows the equi-potential contours and the depletion layers of the RC-LIGBTs at breakdown state. For the conventional RC-LIGBT, the equi-potential contours are sparse and laterally distributed at the surface of the N-drift when the device breaks down at the voltage of 100.2 V. The depletion layer is extended to the lateral N-buffer, which indicates that the N-drift is laterally depleted with maximum BV as shown in Fig. 3(a). For the proposed RC-LIGBT, the device will breakdown at the voltage of 342.4 V, and the equi-potential contours are much denser both at the surface and bulk of the N-drift. The depletion layer is extended laterally to the N-buffer and vertically to the BOX as shown in Fig. 3(b). For device A without TOL as shown in Fig. 3(c), the equi-potential contours are almost the same as the conventional RC-LIGBT. For device B as shown in Fig. 3(d) with the right vertical N-buffer layer, the equi-potential contours are sparser than the proposed RC-LIGBT.

Fig. 3. (color online) Equi-potential contours of (a) the conventional RC-LIGBT, (b) the proposed RC-LIGBT, (c) device A without TOL, and (d) device B with right vertical N-buffer layer.

Figure 4 provides the surface and bulk electric field distributions for the devices. At the cutline y = 0.5 μm as shown in Fig. 4(a), the interface of the surface is formed by SiO2/Si for the conventional RC-LIGBT and SiO2/SiO2 for the proposed. The surface maximum electric field is improved from 2.8 × 105 V/cm for the conventional RC-LIGBT to 6 × 105 V/cm for the proposed RC-LIGBT at the source region, and another peak electric field of 3.2 × 105 V/cm is introduced at the collector region for the proposed RC-LIGBT, thus it exhibits a trapezoidal electric field distribution. Compared to the conventional RC-LIGBT with triangular electric fields in the N-drift, the breakdown voltage is greatly improved. For device A without TOL, the electric field is also triangular distributed as the conventional RC-LIGBT. For device B, the left and the right peak electric fields are 5.7 × 105 V/cm and 2.2 × 105 V/cm, respectively, which are lower than the proposed RC-LIGBT because the vertical N-buffer is designed to the right and less effective to stop the lateral electric field. At the cutline y = 5 μm as shown in Fig. 4(b), the N-drift region is formed by Si/Si for the conventional RC-LIGBT and Si/SiO2 for the proposed, the bulk electric field is improved from 1 × 105 V/cm for the conventional to 2.6 × 105 V/cm for the proposed. Another peak electric field of 2.4 × 105 V/cm is also introduced at the collector region for the proposed RC-LIGBT, thus the bulk electric field is greatly improved.

Fig. 4. (color online) The electric field distributions for the conventional RC-LIGBT, proposed RC-LIGBT, device A, and device B. (a) Surface field at the cutline y = 0.5 μm (illustrated in Fig. 1), and (b) Bulk field at the cutline y = 5 μm (illustrated in Fig. 1).

Figure 5 shows the BV as a function of the doping of the N-drift and N-buffer. At the Ndrift = 1 × 1014 cm−3, the BV of the proposed RC-LIGBT increases significantly with the increase of the N-buffer. However, it will remain constant at Ndrift = 5 × 1014 cm−3 and Ndrift = 1 × 1015 cm−3. More importantly, at the same value of the Nbuffer, the BV will increase when the Ndrift changes from 1 × 1014 cm−3 to 5 × 1014 cm−3, but it will decrease when the Ndrift increases from 5 × 1014 cm−3 to 1 × 1015 cm−3, and the maximum BV is obtained when the Ndrift is 5 × 1014 cm−3. This phenomenon is also observed for the conventional RC-LIGBT, i.e., the breakdown voltage has a maximum value when the Ndrift is 5 × 1014 cm−3.

Fig. 5. (color online) The breakdown voltage BV as a function of the N-drift and N-buffer doping for the RC-LIGBTs.
3.2. Snapback Characteristics

At the forward conduction state, the emitter electrode of the device is connected to the ground VEmitter = 0 V. The gate electrode is added with VGate = 15 V, and positive voltage is applied to the collector electrode as VCollector. The collector current and the voltage curves are shown in Fig. 6. For the conventional RC-LIGBT, with the increase of the VCollector, the conduction mode changes from unipolar mode (conducts with electron at point A) to bipolar mode (conducts with electron and hole at point B) due to the short effects of the N-collector. The inner mechanisms of the snapback are given by the inset picture of Fig. 6. From point O to point A, the electron current is injected from the N+ emitter and it flows laterally through the N-drift to the lateral N-buffer, and finally to the N-collector. The voltage drop VPN for the P-collector/N-buffer junction is lower than 0.7 V at the moment. From point A to point B, the VPN reaches 0.7 V and the P-collector starts to inject holes to the N-drift, thus the current increases abruptly and then the snapback occurs. For the proposed RC-LIGBT, the N-buffer is vertically designed and the length of the electron channel is much longer than that of the lateral N-buffer, thus the VPN can easily reach 0.7 V and the P-collector can inject holes at the collector voltage VCollector = 1.2 V when Ndrift = 5 × 1014 cm−3. Moreover, the snapback is completely eliminated when Ndrift = 1 × 1015 cm−3.

Fig. 6. (color online) Forward conduction characteristics of the conventional and the proposed RC-LIGBT, the inset pictures are for the conventional RC-LIGBT at point A and point B.

Figure 7 shows the influence of the N-buffer doping on the snapback characteristic of the RC-LIGBTs. For the conventional RC-LIGBT, when the electron current flows through the N-buffer, the voltage drop VPN is introduced. However, it is hard to reach 0.7 V because the N-buffer is laterally designed, as illustrated in the inset picture of Fig. 7. When the doping of the N-buffer decreases from 1 × 1016 cm−3 to 5 × 1015 cm−3, the snapback can be suppressed, but cannot be eliminated. For the proposed RC-LIGBT, the VPN is much higher than that of the conventional because the N-buffer is vertically designed, and the snapback can be eliminated when Nbuffer is 5 × 1015 cm−3.

Fig. 7. (color online) The influence of the doping of the N-buffer on the forward conduction characteristic of the RC-LIGBTs; the inset pictures are the conventional RC-LIGBT with lateral N-buffer and the proposed RC-LIGBT with vertical N-buffer.
3.3. Reverse conduction characteristics

At the reverse conduction state, the emitter electrode VEmitter of the device is connected to the positive voltage. The gate electrode and the collector electrode are connected to the ground, i.e., VGate = 0 V and VCollector = 0 V. The devices are working at the diode mode, with the P-body acting as the anode and the N-collector acting as the cathode of the diode. Figure 8 shows the influence of the doping Nbuffer on the reverse conduction of the RC-LIGBTs. For the proposed RC-LIGBT, the N-buffer is vertically designed and acts as the electron current channel, thus the reverse conduction capability will be improved when the Nbuffer increases from 5 × 1015 cm−3 to 1 × 1016 cm−3. However, the reverse voltage drop is 1.78 V at 100 A/cm2 when the Nbuffer = 1 × 1016 cm−3, which is higher than that of the conventional RC-LIGBT with 0.85 V.

Fig. 8. (color online) The influence of the doping Nbuffer on the reverse conduction for the RC-LIGBTs.

Wn and Wp are the widths of N-buffer and P-collector, respectively, as illustrated in Fig. 1. Figure 9 shows the influence of the ratio Wn/Wp on the reverse conduction of the RC-LIGBTs. For the proposed RC-LIGBT, the width Wn of the electron current channel will be improved with the increase of the ratio Wn/Wp, as shown in the inset picture, so the reverse conduction capability will be improved when the ratio Wn/Wp changes from 1:2 to 3:1. The reverse voltage drop is 1.58 V at 100~A/cm2 when the ratio Wn/Wp = 3:1.

Fig. 9. (color online) The influence of the ratio Wn/Wp on the reverse conduction for the RC-LIGBT; the inset pictures are the current distribution for the proposed RC-LGBT.
3.4. Turn-off characteristics

The switching test circuit shown in Fig. 10 is used to investigate the turn-off performance.[13] In the simulation, the diode is idea, the gate resistor Rg is 10 Ω, the stray inductance Ls is 10 μH, the gate voltage Vg changes from 20 V to −5 V, with a frequency of 5 kHz and a 50% duty cycle to control the device’s turning on and off. In addition, the bus voltage is set as 200 V for the proposed RC-LIGBT, and 60 V for the conventional LIGBT (the Vbus is set as 60% of the breakdown voltage).

Fig. 10. The resistance load switching test circuit with Rg = 10 Ω, Ls = 10 μH, and Vbus = 200 V for the proposed RC-LIGBT, Vbus = 60 V for the conventional LIGBT (Vbus is set as the 60% of BV of the devices).

The switching current and voltage waveforms are shown in Fig. 11. The devices are turned off at the same current density 100 A/cm2. The turn-off time Toff is calculated from 90% IC to 10% IC. For the proposed structure, the vertical N-buffer acts as the fast path for extracting the excess carriers in the N-drift, as the results show the Toff are 980 ns and 630 ns for the proposed device with Tox = 22 μm and Tox = 11 μm. The Toff is 310 ns for the conventional RC-LIGBT (Tox = 0). For the conventional LIGBT, the excess carriers disappear only depending on the recombination. The Toff is 1.85 μs at the doping NP-Collector = 8 × 1017 cm−3, and it can be further optimized by the NP-Collector. The Toff is 1.23 μs and 850 ns for the NP-Collector = 5 × 1017 cm−3 and 3 × 1017 cm−3, respectively.

Fig. 11. (color online) Turn-off current and voltage waveforms of the devices under a resistive load switching circuit. JF = 100 A/cm2, Rg = 10 Ω, L = 10 μH, and the Vbus is set as the 60% of BV of the devices.

Figure 12 shows the influence of the ratio Wn/Wp on the switching property of the proposed RC-LIGBT with Tox = 22 μm. The turn-off capability will be improved when the ratio increases from 1:2 to 3:1, and the turn-off time will be reduced from 980 ns to 610 ns due to the enhanced extracting capability of the vertical N-buffer.

Fig. 12. (color online) The influence of the ratio Wn/Wp on the switching property for the proposed RC-LIGBT with Tox = 22 μm.
4. Conclusions

A novel RC-LIGBT with vertical N-buffer and P-collector is proposed. The results show that at the reverse breakdown state, the BV is greatly improved by 340% with the improvement of the surface and bulk electric field in the N-drift region. Moreover, at the forward conduction state, the snapback can be eliminated by increasing the resistance of the vertical N-buffer. However, at the reverse conduction state, the voltage drop is higher than that of the conventional RC-LIGBT, and can be further optimized by increasing the doping of the N-drift, N-buffer, and the ratio Wn/Wp. Additionally, the turn-off property of the proposed RC-LIGBT is superior to the conventional LIGBT, and it can be optimized by the thickness Tox of the trench oxide and the ratio of Wn/Wp.

Reference
[1] Zhu J Zhang L Sun W Du Y C Huang K Q Chen M Shi L X Gu Y Zhang S 2016 IEEE Trans. Electron. Dev. 63 1161
[2] Sun W Zhu J Zhang L Yu H Du Y C Huang K Q Lu S L Shi L X Yi Y B 2015 IEEE Electron. Dev. Lett. 36 693
[3] Zhu J Sun W Dai W Zhang L Liu S L Shi L X Yi Y B Zhang S Sun W 2014 IEEE Trans. Electron. Dev. 61 3814
[4] Son W S Sohn Y H Choi S 2004 Microelectron. J. 35 393
[5] Varadarajan K R Chow T P Wang J Liu R Gonzalez F 2007 IEEE ISPSD 2007 233
[6] Luo X R Wang Q Yao G L Wang Y G Zhou K 2013 Chin. Phys. 22 027303
[7] Napoli E Spirito P Strollo A G M Frisina F Fragapane L Fagone D 2002 IEEE Electron Dev. Lett. 23 532
[8] Zhang L Zhu J Sun W Du Y C Yu H Hua K Q Shi L X 2015 IEEE ISPSD 49
[9] Chen W Z Li Z H Ren M Zhang J P Zhang B Liu Y Hua Q Mao K Li Z J 2013 IEEE ISPSD 265
[10] Chen W Z Li Z H Zhang B Ren M Zhang J P Liu Y Li Z J 2014 Chin. Phys. 23 018505
[11] Byeon D S Chun J H Lee B H Kim D Y Han M K Choi Y I 1999 Microelectron. J. 30 571
[12] Chul J H Byeon D S Oh J K Han M K Choi Y K 2000 IEEE ISPSD 149
[13] Chen W Zhang B Li Z J 2010 IEEE Electron Dev. Lett. 31 467
[14] Zhu J Zhang L Sun W Chen M Zhou M Zhao M Shi L X Gu Y Zhang S 2016 IEEE Trans. Electron Dev. 63 2003
[15] Zhu J Zhang L Sun W Chen M Zhao M Huang X Chen J Qian Y 2017 IEEE Electron Dev. Lett. 64 1187
[16] Sin J K O Mukherjee S 1991 IEEE Electron Dev. Lett. 12 45
[17] Green D W Sweet M Vershinin K V Hardikar S Narayanan E M S 2005 IEEE Trans. Electron Dev. 52 2482
[18] Jiang H P Zhang B Chen W J Li Z J Liu C Rao Z G Dong B 2012 IEEE Electron. Dev. Lett. 33 417
[19] Vemulapati U Kaminski N Silber D Storasta L Rahimo M 2014 IET Circuits Device Syst. 8 168
[20] Chen W Z Zhang B Li Z H Ren M Li Z J 2012 IEEE ICSICT 1
[21] Taurus Medici Davinci User’s Guides, Synopsys 2013